USE OF A COST-EFFECTIVE NEUROHEADSET EMOTIV EPOC FOR PATTERN RECOGNITION PURPOSES

Transcription

1 ISSN International Journal of Computing USE OF A COST-EFFECTIVE NEUROHEADSET EMOTIV EPOC FOR PATTERN RECOGNITION PURPOSES Alekandra Kawala-Janik 1), Jerzy Baranowki 2), Michal Podpora 1), Pawel Piatek 2), Mariuz Pelc 1,3) 1) Faculty of Electrical Engineering, Automatic Opole Univerity of Technology, ul. Prozkowka 76, Opole, Poland, {a.kawala-janik, m.podpora, m.pelc}@po.opole.pl, 2) AGH Univerity of Science and Technology, Faculty of Electrical Engineering, Automatic, Computer Science and Biomedical Engineering, al. Mickiewicza 30, Krakow, Poland, {jb, ppi}@agh.edu.pl, 3) School of Computing and Mathematical Science, Univerity of Greenwich, Old Royal Naval College, Park Row, SE10 9LS, London, United Kingdom, m.pelc@greenwich.ac.uk, Abtract: Application of biomedical ignal for the control purpoe i currently growing interet of reearch ociety. Variou biomedical ignal enable variou control propect. In thi paper application domain of uing electroencephalographic ignal obtained from an inexpenive Emotiv EPOC headet wa decribed. It i alo important to mention the poible implementation of the propoed method on an embedded platform, a it caue ome ignificant limitation due to the little efficiency and low computing power of an embedded ytem platform. The propoed method enable to extend future application of the BCI ytem preented in thi paper and it alo give more teting flexibility, a the platform can imulate variou external environment. It i crucial to mention, that no filtering wa done and that the traditional, tatitical ignal proceing method were in thi work neither ued, nor decribed. Copyright Reearch Intitute for Intelligent Computer Sytem, All right reerved. Keyword: Electroencephalography (EEG), Brain-Computer Interface (BCI), ignal proceing, bio-informatic, control, robotic. 1. INTRODUCTION Implementation of variou biomedical ignal in particular EEG a an information ource applied for the purpoe of external environment control ha become recently growing concern in the cientific world. Application of the EEG ignal a a data ource for Brain-Computer Interface (BCI) enable quick and direct communication between the computer (or any other device) and the brain [1]. Mot of the BCI ytem require expenive equipment with high computing power and it ued complex ignal proceing method, what eliminate the majority of the current olution from their potential implementation on embedded platform. Thi i becaue the analyi of biomedical ignal and in particular EEG i very complex due the preence of variou artifact, which can be both internal and external. The ignal itelf are very enitive to variou diturbance and nontochatic [2-4]. A thee biomedical ignal have non-tationary character, thi may frequently lead to variation of SNR (Signal-Noie-Ratio), where low SNR can caue inufficient decoding accuracy and a a reult affect the overall quality of a BCI ytem [3-5]. The novelty of the olution decribed in thi work relie on application of baic mathematical operation, uch a addition, ubtraction, multiplication and diviion, only. Traditional, ophiticated ignal proceing method were not applied. All experiment carried out for the tudy purpoe were conducted in noiy, imilar to real-life condition [6]. 2. BRAIN-COMPUTER INTERACTION Lat few decade have brought very thorough exploration of the BCI-related field of reearch. BCI ytem can be divided into two main group invaive (with urgical implantation of electrode) and non-invaive (e.g. thoe EEG-baed, where no 25

2 urgical intervention i made). It i alo poible to divide the ytem into the two varietie ynchronou and aynchronou [1]. Any BCI ytem' goal i to record the brain activity in order to manage computer or machine action [7]. In Fig. 1 the main ix tep carried out in a typical BCI ytem were preented, where at the very beginning the cerebral activity i being recorded and then the data i extracted, claified and finally tranlated into command, which enable to control a computer [7]. 3. RESEARCH METHODOLOGY A mentioned above Emotiv EPOC headet wa applied for thi tudy purpoe. Thi i a 14- channel EEG-baed headet [8-10]. It i alo important to mention that mot of the currently available olution require application of complex ignal proceing method, which reult in need of an expenive equipment due to the high computing power requirement. The method preented in thi paper differ from other mot common BCI ytem by uing baic mathematical operation only and doe not require implementation of meaurement equipment with high computing capability [6]. All experiment were conducted in imilar to real life condition, a only thi would enure that the propoed olution could be applicable in product uch a wheelchair on crowded and noiy treet. Twenty healthy, anonymou ubject participated in thi tudy (Fig. 3), which had to imagine appropriate hand movement in accordance with the meage appearing on the computer creen (Fig. 4). which played a role of a viual timulu with all neceary intruction diplayed on it. Fig. 1 Six tep in a typical BCI ytem. The BCI ytem preented in thi work i baed on EEG, where electrode are placed on the urface of the calp, which enable to read ignal generated by the electrical activity of the brain [7, 8]. For thi tudy purpoe a non-invaive, EEG baed Emotiv EPOC headet wa applied. A it wa important to ue uer-friendly, cot-effective, commercial headet (Fig. 2) [6, 8, 9]. Fig. 3 Anonymou reearch participant. Fig. 2 Neuroheadet Emotiv EPOC [9]. Fig. 4 Viual timulu with the tak to be completed by reearch participant. 26

3 The data wa obtained from the two channel only F3 and F4 (Fig. 5). The electrode are placed according to the electrode placement tandard. The data wa recorded during imagery hand movement right (F3 electrode) and left (F4 electrode). Equipment ued during thi reearch i not a typical medical device and therefore doe not regiter ignal with medical preciion. Fig. 5 Electrode placement. The choice of an inexpenive, gaming equipment wa made to preent poible implementation of thi ort of hardware in order to build cheap, but effective Brain-Computer Interface. The Emotiv EPOC conit of 16 electrode, where 14 only are placed on calp and record EEG ignal. The ampling rate of the device i 128 Hz and the bandwidth i between 0.2 and 45 Hz, which i fair enough for the invetigated frequency range [9, 10]. It can alo be uccefully applied for recognition of uer emotion and ha potentially wider ue than traditional clinical electroencephalograph [11]. The neuro-headet ue three type of control EEG, EMG and Gyrocope [12, 13]. It alo ha fewer calp contact than a typical profeional device and potentially le accuracy. Alo very little tudy wa carried out in order to tet it accuracy compared to the traditional EEG [8, 12, 13]. 4. MATHEMATICAL INTERPRETATION OF THE PROPOSED METHOD A mentioned above the method can only be preented with the ue of baic mathematical operation only (1): N 2 M α 1 α = kt p kt + S lf P lf, N ( ) ( ) i i j j k= 0 M l= 0 (1) where t = kt i the dicrete time a k = 0,1,..., N 1, kt and p kt for i i i = 1,...,r, are the dicrete time repreentation of the ith ignal and it pattern (or model depending on ue), repectively, ampled at the frequency 1 F =, where T i the ampling interval, T S j lf and P lf j amplitude pectra of i kt and p kt i are the ingle-ided repectively, with f being the frequency teprelated to (but not necearily equal) to F. The normaliation enure that, p,s,p 0,1 i i j j and that the value of Є alway belong to the [0,1]. It i poible to differentiate two component of ignal for the purpoe of analyi. The weighted difference between the pattern and the ignal i et up for both domain the time domain and the frequency domain. In cae the ignal i of bad quality very noiy, then a a reult it timedomain repreentation may not be very ueful for the reearch purpoe. In thi cae the coefficient hould be et to the value '1' or very cloe to '1', o that only the frequency domain component would be taken into account. Typically a the mot optimal olution the value of the 'α' coefficient hould be et to '0.5', which mean that the both component are equally important. The novelty of the diagnotic (or pattern recognition) approach adopted for the purpoe of thi reearch i an application of a threhold impoed on, which enable to make deciion on the quality of pattern recognition. 5. EXPERIMENTAL RESULTS For thi tudy purpoe all ignal were recorded in a noiy, full of diturbance, environment. Fig. 6 illutrated ignal obtained from the 'F3'-electrode during right-hand imagery movement. The ignal matched. Signal were recorded from two different male, adult ubject Subject 3 and Subject 7. It i alo important to mention, that only raw, unfiltered ignal were proceed. No filtering at all wa done. In Fig. 7 the ame ignal, but in a normalied, caled [0,1] form were preented. In both view it i poible to notice 'peak' preent in ignal, what may be conidered a potential artifact. However the propoed method 27

4 contain feature of mean-quare method. Thi mean that thi method ha attribute of averaging the value and a a reult the eventual 'peak', which may occur in ignal will be eliminated. The ignal alo viualy eem to differ trongly, however it i clearly noticable that they ocilate around imilar value. controller would not be able to proceed the computing. The conducted reearch wa carried out on a mall (for tatitic criteria) group of ubject and therefore the obtained reult could be unreliable, however it could be conider a a preliminary tudy (imilar to: [14, 15]). Analyi of efficiency of the propoed olution for the particular candidate ha not been done at thi tage. A the ophiticated ignal proceing method cannot be ued a it may caue prohibitive computational burden. Alo the device itelf (Emotiv EPOC headet) had ome diadvantage, a it wa not deigned for clinical uage. The obtained ignal did not contain full information unlike it i in cae the ignal are recorded with a typical electroencephalograph. A a reult - the recorded EEG ignal had alo a very low accuracy. Fig. 6 Subject 3 and 7 imagery right-hand movement 'F3'-electrode. 28 Fig. 7 Subject 3 and 7 imagery right-hand movement 'F3'-electrode normalied view. 6. RESEARCH CHALLENGES All the above mentioned numeric procedure were carried out in MATLAB oftware package. The whole reearch wa challenging due to the nontationary nature of biomedical ignal. Alo implementation of the embedded platform ha caued ignificant limitation in chooing appropriate ignal proceing method. A it alo preclude application of advanced neural network, a the traditional embedded platform micro Fig. 8 Scheme of the communication between the particular BCI component. A the current approach i baed on the fat prototyping cheme. Tetbed wa baed on communication between PC and TS-7260 (embedded platform) and between MATLAB and PC. The main aim of thi work wa to build a fully working, tandalone BCI ytem with no need of uing Matlab or PC. The tructure of the BCI ytem fat prototyping tetbed decribed in thi paper wa preented in Fig. 8.

5 In future application it will be deired to contruct a imple embedded ytem baed on a mobile computing platform. Main approach, that i conidered i the automatic code generation directly from MATLAB/Simulink with ue of Simulink Coder and Matlab Coder toolboxe ( [27]. 7. CONTRIBUTION OF THE RESEARCH In thi tudy traditional complex tatitical ignal proceing method were not involved. The novelty of the propoed olution relie on application of the baic mathematical operation. The propoed method i imple, novel and what i the mot important efficient. No filtering wa done a it did not improve the reult (ome of the pilot tudy wa done in order to tet the effectivene of filtering ee: [26]. It i alo important to mention, that the Emotiv EPOC provide wirele USB connector and ha relatively good battery life up to 12 hour work [16]. The ignal recorded with Emotiv EPOC headet are quite noiy [17]. Alo a already mentioned above Emotiv EEG (or EPOC) i an inexpenive, non-invaive, 'off-the-helf' wirele EEG neuroheadet, where the raw data tream i locked. Alo the recording accuracy i low, however it enable ucceful implementation in variou BCI application [28]. All the numeric procedure of thi work were conducted in MATLAB. Adopted tool for ignal proceing could be more ophiticated, although it might led to prohibitive computational burden, in particular in the embedded ytem environment elected owing to the lowcot implementation prerequiite. Alo the implementation of Emotiv EPOC headet had ome diadvantage, a the device wa not ued for clinical application and therefore the accuracy of the regitered ignal wa not very high, however the conducted initial tet proved efficiency and uitability of the implementation of the propoed olution in real-life environment. A it wa mentioned above, equipment ued for the reearch purpoe wa not deigned for clinical ue. It i inexpenive and eay to ue (alo for inexperienced potential uer) [1, 18, 19]. The propoed device became very popular recently among other BCI reearcher due to it intuitive uer interface and low price [18, 19]. Other BCI olution (eg. Khuhaba [21], Voloyak [22] or Cholula [23]) are baed on analyi of variou brain-ignal, uch a α or β unlike the propoed by the author of the hereof paper method, where only the wave are being proceed. Many cientit engaged in analyi of EEG ignal tate that the frequencie of: δ, θ and α are trongly correlated with drowine, fatigue and poor tak performance [29]. The above mentioned method alo require high computing power, what make it impoible to implement in traditional embedded platform. The novelty of the decribed method relie alo on it implicity and lack of traditional tatitic method, applied in other BCI ytem (e.g. [21-23). For the reearch purpoe only two electrode were taken into conideration, and a a reult only two channel have been ued 'F3' and 'F4' [22]. The propoed method' efficiency i quite high 91.7% in cae ignal wa gathered in quiet environment, recorded during left hand movement from the electrode placed on 'F4' poition, which wa upriingly high. However in cae the ame ignal, from the ame electrode were gathered in different condition noiy environment the efficiency dropped to 86.7%. For the ignal (both noiy and quiet environment) obtained during righthand movement and recorded from the 'F3' electrode the efficiency wa the ame 86.7%. Traditional method SSVEP (Steady-State Viual Evoked Potential) or P300 Paradigm ocillate between 69.2% and 100% and require higher computing power [15, 22, 23]. SSVEP BCI olution overall pattern-recognition efficiency (preented in: [15]) wa 84%, which i much lower than the effectivne of the olution propoed by the author of thi paper. Alo both SSVEP and P300-baed BCI require implementation of complex ignal proceing method and therefore would not be uitable for the method ued by author of thi publication [15, 23]. Alo no filtering wa done in thi work, a ome initial tet have proven that in analyi of ignal with limited information data (when the ignal have been hardware-preproceed) filtering 'cut-off' alo the valuable information, what ignificantly decreae the overall ignal proceing performance [6, 10, 24]. Alo variou filtering method uch a adaptive patio-temporal (AST) algorithm are frequently applied. The advantage of uch algorithm i that it reduce the danger of ignal overfitting by contructing a low-pa temporal filter with the implementation of two-parameter Gauian kernel [30]. Uing uch a ophiticated method, which would alo involve ome mathematical modelling i wa impoible to implement on the embedded platform with a very limited computing. To um it all up the novelty of the propoed olution (depite it implicity and limitation) relie on repudiation of the traditional ignal proceing method baed on complex tatitic. The ue of baic mathematical operation enable potential 29

6 implementation of the method in embedded ytem with mall computing power. The method can be alo eaily tranferred into any programming language including 'C' or 'Aembler'. During very thorough literature tudie imilar method have not been found. 8. FUTURE WORK Further reearch will be carried out in three main area. The firt one, conider the algorithmic and conceptual improvement. The ignal proceing algorithm ha a potential for development, with advanced technique. Epecially method of filtering will be reconider with poible application of tatitical filtering via denitogram [25] or Bayeian filtering [26]. Alo pattern recognition and claification algorithm require additional analyi. Currently field i dominated by neural baed approache, which have many drawback. Author conider application of Bayeian claifier. In order to verify any potential advance in algorithm modification erie of additional experiment i needed. Second area for improvement i focued on hardware realiation of the ytem. In particular two baic approache are conidered. The firt one i to contruct a dedicated microcontroler ytem which will be reponible for data proceing and interpretation. Thi approach ha it merit, epecially if true real time requirement are preent. Thi i where automatic code generation hine the mot, a Simulink algorithm can be moothly implemented. Main drawback are limit to computing power and the requirement of creation of dedicated ytem, which in low number i imply not cot effective. Different approach i to ue a mobile embedded platform uch a martphone. In uch ituation actual electronic circuit are limited to ome kind of wirele interface for example for Bluetooth or WLAN. In uch cae, martphone acquire data from the interface and take care of the proceing. What i epecially intereting, becaue of cellular network there i a contant internet acce and the potential for cloud computing. It i currently ued in uch application like Google Voice Search. Third avenue for development i the extenion of the ytem by additional method of biomedical ignal acquiition. In particular peech, EMG and EOG. Potential for integration of uch ytem i unlimited with broad area of application. The further tudy will focu alo on improvement of the ignal-proceing method and application of other bio-ignal in order to extend the poible applicability and ameliorate it effectivene. It will alo involve improvement of the propoed algorithm in order to improve the pattern recognition efficiency. There are alo plan for conduction more experiment with the implementation of other inexpenive, eaily available on market headet in order to obtain more data and to make the propoed method more reliable. A mentioned already above, all numeric procedure were carried out in MATLAB. The further reearch plan involve building a tandalone ytem, where not only EEG ignal will be ued, but alo other bio- ignal uch a peech, EMG and EOG. Reearch will alo be conducted on wider group of ubject in order to make the obtained reult more reliable. Some initial tet (Fig. 9) were already run on adapting the propoed method for EMG ignal recorded during imple finger movement. The efficiency of the method wa in that cae a little bit lower ca. 85%. Alo ome initial reearch regarding imultaneou on-line analyi of EEG and fmri ignal wa already conducted [31-33]. Fig. 9 Sample of the left (top) and right (bottom) index finger movement. 6. REFERENCES [1] Wenjia Ouyang, Kelly Cahion, Vihayan K. Aari, Electroencephalograph baed brain machine interface for controlling a robotic arm, 2013 IEEE (AIPR) Applied Imagery Pattern Recognition Workhop: Sening for Control and Augumentation, (October 23-25, 2013), pp [2] K. Blinowka, J. Zygierewicz, Practical Biomedical Signal Analyi Uing MATLAB, Serie in Medical Phyic and Biomedical Engineering, CRC Pre, Taylor and Franci Book, [3] S. Gao, Y. Wang, X. Gao, B. Hing, Viual and auditory Brain-Computer Interface, IEEE Tranaction on Biomedical Engineering, (Accepted for Publication) (2014). [4] H. I. Suk, S. W. Lee, A novel Bayeian framework for dicriminative feature extraction 30

7 in Brain-Computer Interface, IEEE Tranaction on Pattern Analyi and Machine Intelligence, (35) 2 (2013), pp [5] D. Huang, K. Qian, D.-Y. Fei, W. Jia, X. Chen, Electroencephalography (EEG)-baed Brain- Computer Interface (BCI): A 2-D virtual wheelchair control baed on Event-Related Deynchronization/Synchronization and tate control, IEEE Tranaction on Neural Sytem and Rehabilitation Engineering, (20) 3 (2012), pp [6] Alekandra Kawala-Janik, Michal Podpora, Mariuz Pelc, Jerzy Baranowki, Pawel Piatek, Implementation of an inexpenive EEG headet for the pattern recognition purpoe, 2013 IEEE (IDAACS) 7th International Conference on Intelligent Data Acquiition and Advanced Computing Sytem, (Berlin, September 12-14, 2013), pp [7] Simon Grude, Matthew Freeland, Chenguang Yang, Hongbin Ma, Controlling mobile Spykee robot uing Emotiv Neuro Headet, 2013 (CCC) 32nd Chinee Control Conference, (Xi'an, July 26-28, 2013), pp [8] Hidenori Boutani, Mieko Ohuga, Applicability of the Emotiv EPOC Neuroheadet a a uerfriendly input interface, 2013 (EMBC) 35th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, (Oaka, July 3-7, 2013), pp [9] Official webite of the Emotiv Inc. Company: [10] Emotiv Wikia Page: com [11] R. Bernay, J. Mone, P. Yau, M. Murcia, J. Gonzale-Sanchez, M. E. Chavez- Echeagaray, R. Chritopheron, R. Atkinon, Lot in the dark: Emotion adaptation, 2012 (UIST) 25th Annual ACM Sympoium on Uer Interface Software and Technology. [12] G. M. Poor, L. M. Leventhal, S. K. Bowling, J. Ringenberg, S. D. Jaffee, Thought Cube: Exploring the ue of an inexpenive Brain- Computer Interface on a mental rotation tak, 2011 (ASSETS) 13th International ACM SIGACCESS Conference on Computer and Acceibility. [13] C. Vi, S. Subramanian, Detecting Error-related negativity for interaction deign, 2012 (CHI) SIGCHI Conference on Human Factor in Computing Sytem, ACM. [14] A. Vourvopoulo, F. Liarokapi, Braincontrolled NXT robot: Tele-operating a robot through brain electrical activity, rd International Conference on Game and Virtual Word for Seriou Application. [15] O. Friman, I. Voloyak, A. Graeer, Multiple channel detection of Steady-State Viual Evoked Potential for Brain-Computer Interface, IEEE Tranaction on Biomedical Engineering, (54) 4 (2007). [16] J.-M. Seigneur, The Emotional Economy for the Augmented Human, (AH) 2nd Augmented Human International Conference, ACM, [17] A. T. Campbell, T. Choudhury, S. Hu, H. Lu, M. K. Mukerjee, M. Rabbi, R. D. S. Raizada, NeuroPhone: Brain-Mobile Phone Interface. Uing a Wirele EEG Headet (MobiHeld) 2nd ACM SIGCOMM workhop on Networking, Sytem, and Application on Mobile Handheld, USA (Augut 2010). [18] M. van Vliet, A. Robben, N. Chumerin, N. V. Manyakov, A. Combaz, M. M. van Hulle, Deigning a Brain-Computer Interface Controlled Video-Game Uing Conumer Grade EEG Hardware (BRC) Bioignal and Biorobotic Conference, Brazil (January 2012). [19] A. Luo, J. Sullivan, A Uer-Friendly SSVEP- Baed Brain-Computer Interface Uing a Time- Domain Claifier, Journal of Neural Engineering, (7) 2 (2010). [20] Y. Liu, X. Jiang, T. Cao, F. Wan, P. Un Mak, P.-I. Mak, M. I. Vai, Implementation of SSVEP Baed BCI with Emotiv EPOC (VECIMS) IEEE International Conference on Virtial Environment Human-Computer Interface and Meaurement Sytem, China (July 2012). [21] R. N. Khuhaba, S. Kodagoda, G. Dianayake, L. Greenacre, S. Burke, J. Louviere, A Neurocientific Approach to Choice Modeling: Electroencephalogram (EEG) and Uer Preference, 2012 (IJCNN) the 2012 International Joint Conference on Neural Network. [22] I. Voloyak, Ch. Guger, A. Graeer, Toward BCI Wizard Bet BCI Approach for Each Uer, 2010 (EMBS) the 32nd Annual International Conference of the IEEE EMBS, Argentina, (Augut-September 2010). [23] G. Roa-Cholula, J. M. Ramirez-Corte, V. Alarcon-Aquino, J. Martinez-Carballid, P. Gomez-Gil, On Signal P-300 Detection for BCI Application Baed on Wavelet Analyi and ICA Preproceing, 2010 Electronic, Robotic and Automotive Mechanic Conference. [24] A. Kawala-Janik, Efficiency Evaluation of External Environment Control Uing Bio- Signal, (PhD Diertation), The Univerity of Greenwich,

8 [25] J. Baranowki, P. Piatek, A. Kawala-Janik, M. Pelc, R. J. Anthony, Application of Kernel Denity for Analyi of EEG Signal, Ubiquitou Computing and Ambient Intelligence, Lecture Note in Computer Science, (7656) 2012, pp [26] L.G. Bretthort, Bayeian Spectrum Analyi and Parameter Etimation, Lecture Note in Statitic, (48) 1988, Springer Verlag, New York. [27] P. Piatek, D. Marchewka, M. Pauluk, Automatyczna generacja kodu regulatora dla wbudowanego terownika ukladu magnetycznej lewiacji, Projektowanie, analiza i implementacja ytemow czau rzeczywitego (red. L. Trybu i S. Samolej), WKL, Warzawa, 2011, pp (in Polih) [28] N. Al-Ghamdi, G. Al-Hudhud, M. Alzamel, A. Al-Wabil, Trial and tribulation of BCI control application, Science and Information Conference (SAI) 2013, pp [29] Ch.-T. Lin, Ch.-H. Chuang, Ch.-Sh. Huang, Sh.-F. Tai, Sh.-W. Lu, Y.-H. Chen, L.-W. Ko, Wirele and Wearable EEG Sytem for Evaluating Driver Vigilance, IEEE Tranaction on Biomedical Circuit and Sytem, vol. 8, no. 2, 2014, pp [30] J. Lu, K. Xie, D. McFarland, Adaptive Spatio- Temporal Filtering for Movement Related Potential in EEG baed Brain-Computer Interface, IEEE Tranaction of Neural Sytem and Rehabilitation Engineering, Accepted for Publictation in DOI: /TNSRE [31] X. Hongjing, R. Dan, M. S. Cohen, Coupled bai learning and regularized recontruction for BCG artifact removal in imultaneou EEGfMRI tudie, 2013 IEEE 10th International Sympoium on Biomedical Imaging (ISBI), 2013, pp [32] A. Kawala-Janik, M. Podpora, M. Blazczyzyn, J. Kolanka-Pluka, B. Grochowicz, Combined fmri- and EEGbaed Brain-Computer Interface Pilot Study, XXXVII International Conference on Fundamental of Electrotechnic and Circuit Theory 2014, p. 72. [33] A. Kawala-Janik, M. Podpora, J. Kolanka- Pluka, M. Konieczny, B. Grochowicz, Pilot Study on On-Line Analyi of Biomedical Signal Uing Labview, XXXVII International Conference on Fundamental of Electrotechnic and Circuit Theory 2014, p. 73. Alekandra Kawala-Janik, Ph.D. received M.Sc. in Computer Engineering from the Opole Univerity of Technology in Poland in In 2012 he ha completed doctorate in Computing and wa awarded in 2013 Ph.D. degree in Computing by the Univerity of Greenwich in London. She ha been working a a Lecturer in both Opole Univerity of Technology and Univerity of Greenwich. She ha alo gained ome indutrial experience. Currently he i employed a aitant profeor at the Department of Robotic and Applied Informatic, Intitute of Electromechanical Sytem and Indutrial Electronic, Faculty of Electrical Engineering, Automatic Control and Informatic, Opole Univerity of Technology. Her reearch interet include BrainComputer Interface, Human-Computer Interaction, Biomedical Signal Proceing and Articial Intelligence. Jerzy Baranowki, Ph.D. He obtained MSc in Control Engineering in 2006 and PhD in technical cience in Currently he i an aitant profeor in AGH Univerity of Science and Technology at the Department of Automatic and Biomedical Engineering. Hi main reearch interet are control theory, Bayeian tatitic, numerical method and optimization. Michal Podpora, Ph.D. received B.Sc. (2002) and M.Sc. (2004) degree in Computer Engineering from the Opole Univerity of Technology in Poland, where he wa alo awarded in 2012 (Ph.D. in Control Engineering and Robotic). Until 2010 he wa employed in indutry ector a a developer of reearch oftware and a a EU-fund fundraier. Since 2010 he wa Reearch Department Manager and Project Manager of cofinanced reearch project in indutry. At the ame time ( ) he wa employed a K-12 IT teacher. He i alo Apple Ditinguihed Educator. He ha been employed a aociate profeor at the Faculty of Electrical Engineering, Automatic Control and Informatic, Opole Univerity of Technology ince Hi main reearch interet include machine viion, viion reaoning, cognitive ytem, robotic ytem and ditributed ytem. 32

9 Pawel Piatek, Ph.D. He obtained MSc in Control Engineering in 2000 and PhD in technical cience in Currently he i an aitant profeor in AGH Univerity of Science and Technology at the Department of Automatic and Biomedical Engineering. Hi main reearch interet are embedded ytem, ignal proceing and dicrete control. Mariuz Pelc, Ph.D. i currently employed at the Univerity of Greenwich, School of Computing and Mathematical Science, Smart Sytem Technology Department and alo at the Opole Univerity of Technology, Faculty of Electrical Engineering, Automatic Control and Computer Science. Hi cientific activitie are pread acro variou domain, including automatic control and robotic, modern computer / embedded (control) ytem, oftware ytem a well a computer cluter and parallel proceing. Hi mot recent cientific activity i focued on autonomic and elf-* ytem and mot of all - validation and verification of uch ytem, interoperability iue and certification of elf-* ytem. He ha publihed in total of nearly 50 reearch paper and 2 book. 33